Dakota
Version 6.20
Explore and Predict with Confidence
|
Public Member Functions | |
ReducedBasis () | |
default constructor | |
void | set_matrix (const RealMatrix &) |
const RealMatrix & | get_matrix () |
void | center_matrix () |
center the matrix by scaling each column by its means | |
void | update_svd (bool center_matrix_by_col_means=true) |
ensure that the factorization is current, centering if requested | |
bool | is_valid () const |
const Real & | get_singular_values_sum () const |
const Real & | get_eigen_values_sum () const |
const RealVector & | get_column_means () |
const RealVector & | get_singular_values () const |
RealVector | get_singular_values (const TruncationCondition &) const |
const RealMatrix & | get_left_singular_vector () const |
the num_observations n x num_observations n orthogonal matrix U; the left singular vectors are the first min(n,p) columns | |
const RealMatrix & | get_right_singular_vector_transpose () const |
the num_responses p x num_responses p orthogonal matrix V'; the right singular vectors are the first min(n,p) rows of V' (columns of V) | |
The ReducedBasis class is used to ... (TODO - RWH)
Class to manage data-driven dimension reduction. The passed matrix with num_observations n rows and num_responses p columns contains realizations of a set of responses. The class optionally centers the matrix by the column means. Stores a singular value decomposition of the passsed data matrix X = U*S*V', which can also be used for PCA, where we seek an eigendecomposition of the covariance: X'*X = V*D*V^{-1} = V*S^2*V'